Both the strengths and weaknesses of AI are gradually becoming more and more apparent to those who are actually using it for hands-on applications, as opposed to the irrelevant protests of neo-Luddites and proto-Butlerian Jihadists who posture about how they can totally tell if a given text or track is AI-generated or not - and it is absolutely proven that with very rare exceptions, they cannot - or lament the incipient decline of the human race into mindless hedonists spending their days ensconced in an AI-generated cocoon of micro-personalized entertainment slop.
The only way to really grasp the limits of these tools is to deploy them in the kinds of high-level, high-value work that they're supposed to be able to do with ease, speed and accuracy, because nobody's paying real money to watch robots dance or read a copycat AI-generated essay on Yeats that's tossed moments after being submitted to the professor.
In the real world of value creation, optics don't count, accuracy counts. Nobody cares if the AI chatbot that churned out the Yeats homework hallucinated mid-stream because nobody's paying for AI output that has zero scarcity value: an AI-generated class paper, song or video joins 10 million similar copycat papers / songs / videos that nobody pays attention to because they can create their own in 30 seconds.
So let's examine an actual example of AI being deployed to do the sort of high-level, high-value work that it's going to need to nail perfectly to replace us all at work..
Ian’s been experimenting with AI tools (NotebookLM, Gemini, ChatGPT) for months on various projects, and he recently shared this account with me:
"My experience has definitely been mixed. On the one hand, sort of high level requests like 'identify the major issues raised in the documents and sort by importance' produced interesting and suggestive results. But attempts to find and pull together details on a person or topic almost always had noticeable errors or hallucinations. I would never be able to trust responses to even what I consider straightforward instructions. Too many errors. Looking for mentions of 'drew' in 150 warrants said he wasn't mentioned. But he was, I've gone back and found those mentions. I think the bots read enough to give an answer and don't keep incorporating data to the end. The shoot from the hip and, in my experience, have often produced mistakes. Sometimes it's 25 answers and one glaring mistake, sometimes more basic."
Let's summarize AI's fundamental weaknesses:
1. AI doesn't actually "read" the entire collection of texts. In human terms, it gets "bored" and stops once it has enough to generate a credible response.
2. AI has digital dementia. It doesn't necessarily remember what you asked for in the past nor does it necessarily remember its previous responses to the same queries.
3. AI is fundamentally, irrevocably untrustworthy. It makes errors that it doesn't detect (because it didn't actually "read" the entire trove of text) and it generates responses that are "good enough," meaning they're not 100% accurate, but they have the superficial appearance of being comprehensive and therefore acceptable. This is the "shoot from the hip" response Ian described.
In other words, 90% is good enough, as who cares about the other 10% in a college paper, copycat song or cutesy video.
But in real work, the 10% of errors and hallucinations actually matter, because the entire value creation of the work depends on that 10% being right, not half-assed.
Where AI is most effective, and where it is going to inevitably have a powerful effect on employment is in the creative arts. Artists are uniquely vulnerable to AI because the value of their work has no connection to the amount of labor put into it, the extreme variability of quality produced by human artists, and the absolute unreliability of the artists.
I can assure you that even if AI art cost as much as human artists do, every single publisher would replace most human illustrators in a microsecond simply due to the fact that illustrators tend to be, on average, the most stupid, short-sighted, and unreliable people you can imagine.
I used to wonder why illustrators are never, ever, given advances the way authors are. It took me precisely one hiring experience to find out why, when the first illustrator Arkhaven ever hired took the advance he’d been given to illustrate three issues and promptly vanished after turning in just two pages. And that was just the first in a long line of illustrators a) dropping a job in mid-contract, b) not turning in anything for over a year, c) deciding they didn’t like the genre for which they were most known and specifically hired, d) jumping to the competition, and e) changing their fundamental art style.
So if you ever wondered why it takes so long to finish a comic book project, now you know. The more art required, the more likely it is that the artist who is supposed to produce it will flake somehow before he’s done, and this holds true no matter how well the artist is being compensated.
Meanwhile, in the space of about 30 seconds and for the price of a few dollars per month, I can produce an image that is capable of rivaling the very best that Michael Whelan has ever produced in the course of his long and illustrious career.
Textual AI isn’t quite as advanced, but it is already capable of producing results that are every bit as good as the successful genre authors, even if it doesn’t rise to the heights of the literary greats of yore. In the hands of an experienced author, AI-generated text is effective for fiction, screenplays, and comic scripts alike, as this example from an upcoming supernatural action thriller should suffice to demonstrate.
The car pulled onto a service road between two buildings. The darkness was complete except for the headlights showing chain link and concrete.
"Get out," the driver said. His whole demeanor had changed. The service smile was gone.
"No. Take me back. I'll pay double—triple—whatever you want."
"Get out of the car. Now."
"Please, I don't—I have money. I can get more. Just—"
"GET THE FUCK OUT OF THE CAR!"
The locks clicked open.
Mae's hands shook as she reached for the handle. Run, her mind screamed. The moment you're out, run.
But where? She couldn't even see past the headlights.
She stepped out, bare feet on rough asphalt—she'd left her heels in the car. The ground was still warm from the day's heat. Oil and garbage smells. Somewhere, a dog barked.
The car was already moving before she'd fully cleared the door. She stumbled, almost fell. Tail lights disappeared around a corner, taking the last of the light with them.
Darkness so complete it felt solid.
Mae dropped to her knees, feeling around for her phone. There—a faint glow from under where the seat had been. The screen was cracked but working. Her fingers shook as she dialed 191.
"Emergency services," a woman's voice, professionally calm. "What is your emergency?"
"I need help! A Grab driver just—I don't know where I am. Somewhere in Khlong Toei, I think. Please—"
A man’s hand grabbed her phone and yanked it away.
On the other hand, AI has shown itself to be completely unreliable when it comes to programming, legal research, historical facts, citations, sources, and even simple arithmetic.
Which suggests that unless your employment involves doing nothing or moving information from one place to another, such as human resources, marketing, or government-dictated make-work, it is unlikely that you are less valuable than an AI if you are even remotely capable of performing your job functions like a responsible adult.
AI isn’t going to do sales, although it can sit there and take orders as well as any useless account manager. AI isn’t going to make difficult corporate decisions, although it can provide a variety of accountability-free opinions as well as any useless vice-president. And AI isn’t going to fix your toilet, move your refrigerator, or take your dog out for a walk, even if it can mow your lawn.
Whereas on the creative side, it is already apparent that those who embrace and utilize the new AI technologies to enhance their own abilities will tend to win out over those who don’t, just as those who took the time to master Amazon’s A9 algorithm and adjust their independent publishing accordingly beat out those of us who didn’t.
Re. the flakiness of illustrators, I won my current gig in part because the other people who were being interviewed either completely disappeared in the early stages of vetting, or just weren't up to the scope of the work. The problem with artists, in my experience, is that too many of them lack the most basic grounding in essential life skills like showing up and sticking with the process until it's finished. Perhaps similarly, learning to work with AI is too far beyond the scope of their abilities?
Best thing about AI is I can go about my work without calling some IT nerd and listen to him explain how bootloaders work when I really just need to turn something off. Gamechanger.